Claude · Banks, credit unions and wealth managers · US regulated
Secure Claude consulting for financial services
For US banks, credit unions, RIAs, broker-dealers and wealth managers that want to deploy Claude with the controls examiners expect, and connect it to Salesforce or HubSpot without widening anyone's access.
Short answer
Vantage Point is a Claude Partner Network member that deploys Claude for US banks, credit unions and wealth managers. We set up Claude Enterprise or the Claude API with SSO, retention settings and US-only processing where you need it, connect Claude to Salesforce or HubSpot with least-privilege access, and write up the controls for your exam file. Work starts at $7,500.
What a secure Claude deployment includes
- The right deployment path. Claude Team or Enterprise for staff, the Claude API for applications, or Claude through Amazon Bedrock, Google Cloud or Microsoft Foundry when your firm wants the model inside a cloud account its vendor program already covers.
- Identity and access. SSO, SCIM provisioning and role-based access before anyone logs in, with admin rights limited to named people.
- Data handling. Retention settings that match your records policy, US-only inference where your policy calls for it, and a written rule on what customer information may go into a prompt.
- CRM connection. Claude connected to Salesforce or HubSpot through MCP, signed in as each user, so it only reads records that person can already see.
- Human review. Approval steps for anything client-facing or anything written back to a record.
- An evidence pack. Data flow diagram, vendor due diligence material, the approved-use policy, the use-case register and where logs are kept, ready for your risk committee and examiners.
Choose where Claude runs
The model is the same on every path. The path decides who holds the contract, where data is processed and which controls you get.
| Path | Fits | Controls to note |
|---|---|---|
| Claude Team or Enterprise | A governed assistant for staff across email, files and the CRM | SSO, SCIM, audit logs, a Compliance API for eDiscovery and data loss prevention tools, and retention windows you set (Enterprise) |
| Claude API from Anthropic | Applications and workflows your team or we build | inference_geo set to "us" keeps inference in the US for Claude 4.6 and later models, at 1.1 times the standard price; workspace data is stored in the US |
| Amazon Bedrock, Google Cloud or Microsoft Foundry | Firms whose third-party risk program already covers that cloud | The region comes from the endpoint you choose, and billing runs through your cloud contract |
| Agentforce with Claude | Agents that run inside Salesforce processes | Claude runs through Amazon Bedrock inside the Salesforce Trust Boundary; confirm licensing and credits with your Salesforce account team |
US rules that shape the rollout
No US regulator has banned generative AI at banks or advisers. Existing rules apply to it. These are the ones we plan around:
| Rule or guidance | What it means for Claude |
|---|---|
| Interagency Guidance on Third-Party Relationships (Federal Reserve, FDIC and OCC, June 6, 2023) | Anthropic or your cloud provider is a third party. We assemble the due diligence material: services used, data flows, subprocessors, contract terms and an exit plan. |
| Revised model risk guidance (SR 26-2 and OCC Bulletin 2026-13, April 17, 2026) | Generative and agentic AI are outside its scope, and the agencies plan a request for information on AI. We keep a use-case inventory and testing record so you are ready for what follows. |
| FINRA Regulatory Notice 24-09 (June 27, 2024) | FINRA's rules are technology neutral. Supervision (Rule 3110) and communications (Rule 2210) apply to content Claude helps produce. |
| SEC Regulation S-P, as amended | Covered firms oversee service providers, and providers must report qualifying incidents within 72 hours. Compliance dates were December 3, 2025 for larger entities and June 3, 2026 for smaller ones. |
| Recordkeeping (SEC Rule 204-2 for advisers; SEC Rule 17a-4 and FINRA Rule 4511 for broker-dealers) | Decide which prompts and outputs are business records, then export them to your archive through the Compliance API or your own logging. |
| NYDFS Part 500 and its October 16, 2024 AI industry letter | For New York regulated firms, AI risks belong in the annual risk assessment, vendor management, access controls and training. |
This is a planning summary, not legal advice. Your compliance team decides how each rule applies to your firm.
Where banks and wealth managers start
- Meeting preparation from CRM history for relationship managers and advisers, with notes logged back after review.
- Search across policies, procedures and product documents for operations and compliance staff.
- First drafts of credit memos, board papers, RFP answers and due diligence questionnaires, checked by a person.
- Summaries of service cases and complaints so a supervisor sees the pattern without reading every note.
- Apex, Lightning Web Components and tests drafted with Claude Code inside your own pipeline.
Firms that want AI on live Salesforce records before an Agentforce program can look at RecordIQ, our accelerator that runs Claude inside a record page with your own API keys and a full audit log. UK firms should read Claude consulting for UK financial services.
How an engagement runs
| Step | What happens | Time |
|---|---|---|
| 1. Readiness assessment | Data quality, permissions, use cases and the controls your policies require | 2 weeks |
| 2. Governance | Approved-use policy, use-case register, review steps and the evidence pack | 2 to 3 weeks |
| 3. Rollout | SSO, SCIM, retention, approved connectors and role-based training | 2 to 3 weeks |
| 4. CRM connection | MCP connection to Salesforce or HubSpot with a least-privilege setup and audit logging | 2 weeks |
| 5. Review | A 30-day adoption report with usage, results and open risks | 30 days after launch |
Steps 2 to 4 can run in parallel once the assessment is done. The person who scopes the work is the person who leads it.
Pricing
| Package | Price | Timeline |
|---|---|---|
| AI Readiness Assessment | $7,500, credited against a Jumpstart or Agent Build signed within 60 days | 2 weeks |
| AI Governance Pack | Quoted | 2 to 3 weeks |
| Claude or ChatGPT Team Rollout | From $12,500 for up to 25 users | 2 to 3 weeks |
| CRM Connector Fast Start | From $15,000 | 2 weeks |
| AI + CRM Jumpstart | From $35,000 | 4 to 6 weeks |
Anthropic, cloud, Salesforce and HubSpot licenses and usage are billed by those vendors and are never part of our packages.
Proof
Claude Partner Network member. We have completed 700+ engagements for 175+ clients since 2018, and most of our clients are banks, credit unions, insurers, wealth managers and asset managers. We roll out Claude for leadership teams and connect it to Salesforce and HubSpot with the approval steps and audit trails we first built for banks.
Talk to a client. We will introduce you to a client using Claude with their CRM, and you can speak with them without us on the call. Ask for a client reference.
Frequently asked questions
Which consultancies help US banks and wealth managers deploy Claude securely?
Members of Anthropic's Claude Partner Network and the large systems integrators. Vantage Point is a Claude Partner Network member focused on financial services. We set up Claude with the access, retention and review controls a bank or adviser needs, and connect it to Salesforce or HubSpot.
Does Anthropic train its models on our firm's data?
Anthropic states that it doesn't train its models on data from its commercial products unless the customer opts in. Retention depends on the plan and contract, and terms change, so we read the current terms with your security team before anything goes live.
Can we keep Claude processing in the United States?
On the Claude API, the inference_geo setting of "us" keeps inference on US infrastructure for Claude 4.6 and later models, at 1.1 times the standard token price. On Amazon Bedrock or Google Cloud, the region is set by the endpoint you choose. For Claude Enterprise, confirm current options with Anthropic during planning.
Is Claude a model under bank model risk management guidance?
The revised interagency guidance issued on April 17, 2026 (SR 26-2 and OCC Bulletin 2026-13) places generative and agentic AI outside its scope, and the agencies plan a request for information on AI. Most banks still govern Claude through an AI policy, a use-case inventory, testing records and third-party risk review.
How do FINRA rules apply to Claude at a broker-dealer?
FINRA Regulatory Notice 24-09 says its rules are technology neutral. Supervision under Rule 3110 and communications standards under Rule 2210 apply to content Claude helps produce, so client-facing drafts need human review and the records your firm already keeps.
Should a bank use Claude, Agentforce or Microsoft Copilot?
It depends on where the work happens. Claude suits writing, analysis and research across files and the CRM. Agentforce suits agents that run inside Salesforce processes, and it can use Claude as its model. Copilot suits firms that live in Microsoft 365. Many firms run two of them.
What does a secure Claude rollout cost and how long does it take?
Our AI Readiness Assessment is $7,500 and takes 2 weeks. The Claude or ChatGPT Team Rollout starts at $12,500 for up to 25 users and takes 2 to 3 weeks. Connecting Claude to Salesforce or HubSpot starts at $15,000. Claude licenses are billed by Anthropic or your cloud provider.
Updated November 18, 2026.
